Six Sigma the hpye and beyond
12:03 AM
Six Sigma : The Hype and Beyond
Ever since Motorola implemented the Six Sigma standards of quality, it has become fashionable in business circles to talk about it. Many consultancy firms successfully hyped it and also encashed on it. What is so magical about the figure 3.4 defects per million? Is it such a sacrosanct figure or just an arithmetical gimmick? Is Six Sigma a hype or is there any true benefit to be derived from it? This article examines and tries to demystify these issues.
Some time back when the famous Dabbawalas of Mumbai were presented with an award for achieving Six Sigma, the poorly educated dabbawalas were heard asking each other ye sigma kya cheeze hai? (What is this `Sigma'?) Even though the answer is elementary for statisticians, there are glaring gaps in its understanding among the practitioners and even the self-appointed `quality consultants' . "Many of them claim expertise in Six Sigma when they barely understand the tools and techniques and the Six Sigma roadmap." Hence, I start by a quick overview of the fundamentals. When any product is made in large quantities, even the most sophisticated machine will not be producing `exactly' identical products. There are bound to be slight variations, however small it might be. For example, if the product, say a cellphone, is claimed to weigh just 80 gms, individual cellphones produced may actually have weights like 80.03 gms, 79.98 gms, 80.05 gms, etc. Of course, the average weight of many such products made over a period of time will be extremely close to the target value of 80 gms, if not 80 gms itself. The variation of actual values from the target value is measured by sigma, a Greek symbol, which stands for the standard deviation. Smaller the value of standard deviation or sigma, the closer the product values are to the target and larger the value of sigma, the wider the values are scattered around the target value. Hence, the attempt should be to reduce the value of sigma to the minimum possible. Slight variations in individual values, which are unavoidable as explained earlier, are called random variations. But if there exists any specific cause, like malfunctioning of the machine, material defects, operator fatigue or mistake or following the wrong method, etc., it will give glaring deviations to the average of the values got each day or significant fluctuations of individual values. Hence, to keep a track of the quality, each day a sample of products is taken and their average is noted. The average so got should ideally be the target value itself, or, if not, `acceptably close' enough to it. Similarly, the variations in values of individual products produced in a day should also be within `acceptable limits'. Now the big question is, what is this so called `acceptable limit'?
From the rules of statistics, it can be said that the value of `almost all' the individual products produced by a controlled process will fall within three times sigma on either side from the target. Hence, it should be aimed to limit the sigma of variations in individual values within one-third of the permissible or acceptable limits. For example, let us assume that a small increase or decrease of 0.6 gm in the weight of the cellphone from the target value of 80 gms will not create an adverse impact on the customer satisfaction. If so, we can say plus or minus 0.6 gm is well within the tolerance limit of the customers. Hence, we should try to contain the variations in such a way that sigma is just one-third of the tolerance of 0.6 gm. In short, sigma should be a maximum of 0.2 gm. Therefore, if we keep the average weight of cells produced over the days at 80 gms and the weight of individual cells vary well within a maximum value of plus or minus 0.6 gm only, we can be reasonably sure that `almost all' the products produced are good or acceptable to the customers. Here, since the tolerance limit is three times sigma more or less from the target value, the process can be said to be in three sigma control.
Why Six Sigma?
It was explained earlier that if a process is in three sigma control, it will be producing `almost all' good individual products and almost nil defective products. A defect is defined as anything, which could lead to customer dissatisfaction. Now the question is what is meant by `almost all'? Statistically, almost all at three sigma level of quality means actually 99.73% good. Or only 0.27% of the products will be defective. Now the question is can 0.27% be allowed to be defective? In olden days this was acceptable, but due to intense competition, this level of defects became unacceptable in highly competitive fields. The aim shifted to `Zero Defect'. In fact, in large scale operations, 0.27% defect is simply unacceptable. Let us realize that 0.27% defective, though appearing to be negligible, means nearly three defects or failures in a thousand products. In other words, it is 2,700 Defects Per Million Opportunities (DPMO). Translated to reality, it means an air accident every day at Mumbai airport, where nearly 400 planes land per day. Clearly, this is absolutely unacceptable.
The solution for reducing the defect rate lies in reducing the variations in the values of the products from the target to the minimum, so that `all' the products are within the tolerance limit. For this to happen, the defect rate has to be brought down drastically from 0.27%, which is 2,700 per million, to 2 or 3 per million. That is, reducing defects to a thousandth of what is achieved in a three sigma controlled process. The variations in individual values of sigma measure, which was contained within a third of the tolerance limit has to be brought down to a fourth, fifth and ultimately to a sixth of the tolerance limit. So, the journey of improving the quality has to start from three sigma control towards four, five and six sigma control. It was seen that the conventional three sigma level produces 99.73% good ones and 0.27% bad or defectives or 2,700 defects per million. The corresponding limits for defectives in 4, 5 and Six Sigma control will be respectively 63, 0.57 and 0.002 defects per million. The last figure is expressed as 2 per billion. Thus, the fundamental objective of Six Sigma is the progressive improvement of quality by reducing the percentage of defective products to the levels shown above.
The Facts
Many readers must now be thinking loud, "but Six Sigma means 3.4 defects per million". Let us see what this figure of 3.4 defects per million is. This was the figure achieved by Motorola by their Six Sigma implementation initiative under Bill Smith. It was the pioneering attempt by any company in the world to achieve higher standards of quality than a mere 3 Sigma. Hence, all the credit for such initiative must go to Motorola and Bill Smith, no doubt. But the figure of two defects per billion assumes that the long-term average value is exactly at the target value. "According to the Six Sigma philosophy, process rarely stays centered - the center tends to shift above and below the target." If we assume that the center shift is 1.5 Sigma, which means, the actual average value got may be more or less by up to 1.5 Sigma from the target value set for the process. Put it simply, as in our earlier example of weight of the cellphone, if the target aimed at is to have a weight of 80 gms, the average weight of batches of production may have an average weight of 80 gms plus or minus 1.5 Sigma. This was the value assumed by Motorola. Allowing or assuming that the average weight of a batch of products, (and not the individual weights) may be varying upto 1.5 Sigma about the target value set, and containing the individual weight variations within Six Sigma from the target value, Motorola assures their customers that there will be a maximum of only 3.4 defects per million. Now what is so sacrosanct about 1.5 Sigma, one might ask. Why not 1 sigma or why not 2 Sigma? "The magnitude and type of shift is a matter of discovery and should not be assumed ahead of time. None of the case studies in the literature have indicated a shift as great as 1.5 Sigma", opines Dale H Besterfield. According to DH Stamatis, "The automotive industry recognized the concept in the mid 1980s evaluated it and deemed it unacceptable. In fact, the original work of Six Sigma was based on only a few empirical studies of a single process."The statistical aspects of Six Sigma tell us that the variability of individual values of items supposed to be identical should be very low and at the same time, the average of many such batches produced should be centered at or very near the target value (and not 1.5 sigma away from it). Going by the theories of statistics, the defect level of 3.4 per million actually achieved by Motorola corresponds to only 4.5 Sigma level properly centered and not Six Sigma level. Further, these concepts are not anything new, but advocated by quality gurus like Shewhart, Deming, Juran and Taguchi. According to James Harrington, "Six Sigma was simply a TQM process that uses process capability analysis as a way of measuring progress."
In a recent interview, a candidate was asked: "What is the defect rate in the product made or services rendered by a company certified as a Six Sigma company? He answered that the defect rate in a six sigma certified process is no more than 2 defects per billion products. "Is it billion or million?" was the follow-up question. "It is billion" the candidate confidently replied. "No" was the unanimous answer from the interview board. "It is 3 or 4 per million", they asserted. One of them gave the `exact value' - "It is 3.4 defects per million to be exact." The defect rate for a process cannot be estimated unless we specify two factors, the sigma level of control and the assumed shift of the average value from the target value. Table 1 gives the defect rate for processes under three, four, five and six sigma, with the average value of a batch of products, shifted from the target value by 0, 0.5, 1, 1.5 and 2 Sigma. 
From Table 1, it is clear that a question as to what is the defect rate in a process under Six Sigma itself is unanswerable. The answer depends on what is the permitted or assumed shift of center too. If nothing is mentioned about the shift, one must assume that there is zero shift as default value as per the theories of statistics and should not take the value of 1.5 Sigma assumed by Motorola. Hence, for the interview question, which itself was wrong or at least ambiguous, the answer given by the interviewee as 2 per billion was the best possible. "Motorola's design limit of six sigma with a shift of the process off the mean by 1.5 Sigma, gives 3.4 defects per million. If the mean is exactly at the center, then two defects per billion are expected." In fact, a Six sigma process can give defects from 2 per billion at its best to any level of defects, depending on the shift of the average value from the target value.
The Hype
One can see many books stating this value of 3.4 defects per million as the value of Six Sigma quality. However, standard texts and journals give better explanations. "A process that is in six sigma control will produce no more than two defects out of every billion units. Often this is stated as four defects per million units, which is true if the process is only running somewhere within one sigma of the target specification. " That leads us to the next question, why a vast majority of managers and practitioners believe and even argue that Six Sigma `means' 3.4 defects per million? Credit or discredit for this goes to the hype created by Motorola, GE and the various quality consultants and their consultancy firms. Although GE wasn't the first firm to adopt Six Sigma, their effort and achievement was the most visible. Also, Six Sigma didn't receive serious publicity until GE's Welch made headlines in 1998 with reports of his company's $350 mn in Six Sigma-related savings. "..Motorola' s Robert Galvin came up with it (Six Sigma) and breathed life back into the company, bagging a Baldrige Award in the process. Larry Bossidy rebooted Allied Signal with it and then sold General Electric's Jack Welch on it. GE then made Six Sigma front-page news. Next, Mikel Harry, one of the original Motorola Six Sigma gurus, packaged it, hyped it, tried to trademark it, put a cowboy hat and spurs on it, and partnered (and then un-partnered) with American Society for Quality (ASQ) on it, all the while making a lot of money from it. Once GE reported millions of dollars (which later grew into more than $1 bn) in Six Sigma-related savings. It didn't take long for other large companies to take up the Six Sigma torch." That says enough for the hype created. Other comments include, "Personally, it is given way too much `hype' in my opinion,"and "There is an overselling of Six Sigma by too many consulting firms."
Again, to quote Vivek Sharma, "The buzzword-loving quality field has a new emperor, but does he have any clothes on? `Making fewer bad parts' is a good idea, but is hardly a revolutionary one, or even a new one, at that … but while we are shooting for a number, why not aim for zero bad parts, or, more precisely, the Zero Quality Count (ZQC) system by Shigeo Shingo of Toyota fame? Make `no bad parts' seems to be a better goal than saying, `make a couple of bad parts per million.' In fact, there was a joke going round about this at IBM: Apparently, the computer giant IBM decided to have some parts manufactured in Japan as a trial project. In the specifications, they set out that they will accept only three defective parts per 10,000 and same also shall be replaced subsequently. When the delivery came in, there was an accompanying letter. "We, Japanese people, had a hard time understanding North American business practices. But the three defective parts per 10,000 have been separately manufactured and included in the consignment. Replacements are also included in this consignment itself, to save time. Hope this pleases you."
The assumption of 1.5 Sigma shift does not make any sense at all. It is nothing but an exercise in arithmetic, just a gimmick. From Table 1, it can be seen that this particular issue leads to confusion and erroneous defect figures. It also makes six sigma highly subjective. "For example, Motorola conveniently `unaccounts' for a 1.5 long-term drift, and still calls what they do as Six Sigma, though it is closer to 4.5 Sigma." Hence, straightforward measuring units like defects per million or process capability indices are certainly more objective and unambiguous. "One of the major ruses of the Six Sigma `institution' is that it pretends to be a crucial new quality tool, while in reality it has not introduced even one original tool to the quality field. The truth is that, Six Sigma, does not have statistical techniques of its own, and just plagiarizes Statistical Process Control (SPC). All of these approaches to quality improvement are effective techniques, but they are not new, and definitely were not invented by the Six Sigma institution. Brand marketing has come to the quality field; don't fall for the `new and improved' syndrome." Another warning: "Beware of the hype. Six Sigma doesn't solve all problems and it shouldn't be applied in all situations."
Conclusion
All the foregoing discussion does not mean that one should not go in for Six Sigma quality implementation and that it does not give any benefits to the company implementing it. Going for Six Sigma implementation has many benefits. First of all, the whole organization focuses its attention on quality and customer satisfaction aspect of every process. Involving the employees certainly has motivational benefits. The creativity and problem solving skills of one and all are pooled for organizational benefit. Many traditional processes get a critical relook for betterment. "Six Sigma strategy places a clear focus on achieving measurable and quantifiable financial returns to the bottom line of an organization. No Six Sigma project is approved unless the bottom line impact has been clearly identified and defined." Also, "Six Sigma Certification can be a great investment when it is used in the right way and actually needed by the people who seek it… There are infinitely more businesses that will benefit from the use of Six Sigma, than those that won't. When using the process effectively, results can include improved processes, increased customer satisfaction, and increased profitability. " "If it's a measurable, methodical process that you are trying to improve in order to get bottom line results, Six Sigma might be the ticket." On the other hand, "if you or your company is looking for a `quick fix' that will lead you down the yellow brick road; all of the Black Belts in the world couldn't really help you. However, if you have long-term goals for your company and understand the importance of customer satisfaction, Six Sigma is definitely the tool of your choice."
Though the tools are not revolutionary, many companies which have implemented the system say it was worth the effort. It's not quality for quality's sake, but aims at a return on investment, and hence we may conclude Six Sigma has its benefits.
|
Post a Comment